Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Some spectroscopic tools
Marcella Iannuzzi
Department of Chemistry, University of Zürich
4th CP2K tutorial: Enabling the Power of Imagination in MD Simulations August 31- September 4 2015, Zürich
http://www.cp2k.org
http://cp2k.berlios.de
Electron Density: Cube File
2
n(r) =�
µ�
Pµ��µ(r)��(r)��
µ�
Pµ��̄µ�(R) = n(R)Valence density on regular grids
Cutoff might be too small for high resolution close to the nuclei (all electrons)
i(r) =X
µ
Cµi'µ(r) !X
µ
Cµi'̄(R) = i(R)
-Quickstep- TOTAL DENSITY 8 0.000000 0.000000 0.000000 54 0.349949 0.000000 0.000000 54 0.000000 0.349949 0.000000 63 0.000000 0.000000 0.362827 5 0.000000 9.448631 9.448631 11.338357 5 0.000000 9.448631 9.448631 14.683172 1 0.000000 11.322313 9.448631 13.010846 1 0.000000 7.574948 9.448631 13.010846 1 0.000000 9.448631 11.416848 15.778669 1 0.000000 9.448631 7.480413 15.778669 1 0.000000 9.448631 7.480413 10.242860 1 0.000000 9.448631 11.416848 10.242860 0.16324E-08 0.14425E-08 0.13016E-08 0.12075E-08 0.11584E-08 0.11533E-08 0.11920E-08 0.12755E-08 0.14051E-08 0.15832E-08 0.18123E-08 0.20955E-08 0.24355E-08 0.28348E-08 0.32950E-08 0.38170E-08 0.44000E-08 0.50422E-08 0.57400E-08 0.64880E-08 0.72795E-08 0.81060E-08 0.89579E-08 0.98243E-08
B2H6
Spin Density
3
Spin polarized DFT calculations:
n(�)(r) =X
µ⌫
P (�)µ⌫ 'µ(r)'⌫(r)n(↵)(r) =
X
µ⌫
P (↵)µ⌫ 'µ(r)'⌫(r)
�nspin(r) = n(↵)(r)� n(�)(r) ! �nspin(R)
H3C-CO-NH2-CH3
•First •Prev •Next •Last •Go Back •Full Screen •Close •Quit
Valence Electron DensityExample:H3C � CO �NH2 � CH3
n(R)
(H3C-CO-NH2-CH3)+
•First •Prev •Next •Last •Go Back •Full Screen •Close •Quit
Spin DensityExample:[H3C � CO �NH2 � CH3]+
�nspin(R)
Not localized Spin Density
4
Dissociation of HNO3 catalysed by graphite
•First •Prev •Next •Last •Go Back •Full Screen •Close •Quit
Spin Density 6= Localized ChargeDissociation: HNO3 + graph.! NO·2 + OH ·/graph.
•First •Prev •Next •Last •Go Back •Full Screen •Close •Quit
Spin Density 6= Localized ChargeDissociation: HNO3 + graph.! NO·2 + OH ·/graph.
Density Difference
5
Changes in the electronic density due to interactions, e.g., molecule adsorbed on substrate
Hexa-iodo Cyclohexa-m-phenylene (CHP)Molecules synthesized by M. Kastler / K. Müllen MPI MainzFigures: Marco Bieri at al., Chem Com. 45,7 (2009)
Marco Bieri at al., JACS accpeted (2010)
CHP on Cu(111) @ RT
DFT
On Ag(111), Au(111) and Cu(111) the iodine
dissociates from the CHP at temperatures
below 300 K.
This is in agreement with experiences from
the adsorption of hexo-iodo-benzene on these
surfaces.
I 3d
! !
!"#$%&'()*+,(-+$"%
.'/'0',
.'1'0')2(-3
4$55&,&%-&'$%'#&%*$+$&*'672")(2'*8*+&9:';'6$*"2(+'*
STM images
6
z : nb(X,Y, z) e�2kR0�
�(X,Y,z)
Tersoff-Hamann approximation to mimic the iso-current topography
nb(r) =X
i:"i2[Ef�Vb:Ef ]
"X
µ⌫
C⇤µiC⌫i'µ(r)'⌫(r)
#! nb(R)
Find height at constant energy projected density
CHP at h-BN/Rh(111) NanomeshAnnealing @ approximately 120° C for 150 sec
CHP at h-BN/Rh(111) NanomeshAnnealing @ approximately 120° C for 150 sec
occupied states unoccupied states
Vb=-1310 mV Vb=+1480 mVCHP on hBN/Rh (5 eV)
Position operator for isolated systems
7
One body operator
X̂ =NX
i=1
xi
Expectation value
hXi = h |X̂| i =Z
x n(x)dx
Gauge invariance
hXiR = h |X̂ +R| i = hXi0 +RZ
n(x)dx = hXi0 +RZ
Position Operator with PBC
8
Expectation value of the position operator
R̂ =X
i
r̂i hRi = h |R̂| i =Z
rn(r)dr
Wavefunctions are periodic, result of an operator has also to be periodic
(r) = (r+ L) R̂ (r) 6= (R̂+ L) (r+ L)Periodic Position Operator
hXi =L
2⇡Im lnh | ei
2⇡L
X̂ | i
• many-body operator!
• correct asymptotic value L !1; N !1; N/L ! n0
• Polarization
Pel = limL!1
e
2⇡Im lnh | ei
2⇡L
X̂ | i
Berry PhaseMany-body periodic position operator (1D)
Resta, R. (1998). Quantum-Mechanical Position Operator in Extended Systems. Physical Review Letters, 80(9), 1800–1803.and more by R. Resta …..
(3D)
Periodic Position Operator
hXi =L
2⇡Im lnh | ei
2⇡L
X̂ | i
• many-body operator!
• correct asymptotic value L !1; N !1; N/L ! n0
• Polarization
Pel = limL!1
e
2⇡Im lnh | ei
2⇡L
X̂ | iElectronic polarisation
Polarisation
9
2⇡
L= G1
| i = AY
i
Y
s
qs,i(r)
Many-body wavefunction (Bloch orbitals)
New set of Bloch orbitals
̃
qs,m(x) = e�i 2⇡L x
qs,m(x)
hXi = � L2⇡
Im lnh | ̃i
Expectation value from overlap of determinants
S↵ij =
Z i(r)e
iG↵1·r j(r)dr P↵ =
2e
G↵1Im ln [detS↵]
= determinants of overlap of orbitals
hXi = � L2⇡
Im lnh | ̃i = � L2⇡
Im ln detS
Boys spread of the orbitals through a 2-el operator
With PBC, localize equivalent to minimize
Find the unitary transformation
Iterative procedure (parallel Jacobi rotations)
Centre of the charge distribution of the rotated orbital
Localised Orbitals
10
Ω =1
2π
∑
s
∑
i
ωs(1 − (|zsi|2)) zsi =
∫dr eiks·r|ψi(r)|
2
ψ̃i(r) =∑
j
Uijψi(r)∂Ω
∂Uij= 0
R.Resta, Phys. Rev. Lett., 82 370 (1999); G.Bergold et al, Phys. Rev. B, 61 10041 (2000)
⌦ =X
i
h i i|(r1 � r2)2| i ii
KS MLWFhrsii =
Ls2⇡
Im ln zsi
Wannier Centers (3D)
11
For a generalised 3D box h, for each maximally localised Wannier orbital
zsi = deth
Zdr eiks·r| i(r)|2 rsi = �
X
t
hst2⇡
Im ln zti
•First •Prev •Next •Last •Go Back •Full Screen •Close •Quit
Wannier CentersExample:H3C � CO �NH2 � CH3
Molecular dipole moment from Wannier centres
µWs = eX
i
rsi = �eX
i
X
t
hst2⇡
Im ln zti = �eX
t
hst2⇡
Im lnY
i
zti
IR spectra from dipole moment autocorrelation function
•First •Prev •Next •Last •Go Back •Full Screen •Close •Quit
IR SpectraIR absorption coefficient ↵(!) can be calculated from
↵(!) =4⇡ ! tanh(�~!/2)
~n(!)cV
Z 1
�1dt e�i!thP (t) · P (0)i ,
• V: volume of supercell• T: temperature, � = 1/kBT• n(!): refractive index• c: speed of light in vacuum• quantum effect corrections: factor tanh(�~!/2)More sophisticated treatments of quantum effects are also available. Theyconsist of replacing the classical correlation function with a more involvedprocedure.
Materials Science Research
12
sample
measurementtheoretical model
interpretation
Spectroscopies
13
Spectroscopiesmethods to study the properties of matter (atoms,
molecules, solids) by investigating its interactions with particles (photons, electrons, neutrons, ions)
Various Spectroscopic Methods
14
J. Phys.: Condens. Matter 24 (2012) 124106 M-P Gaigeot et al
properties of the oxide surfaces and pKa of surface groups isreported in section 4. The vibrational density of states togetherwith an assignment of the bands in terms of organization ofthe water molecules at the interface is discussed in section 5.Discussion and perspectives are presented at the end of thispaper.
2. Methods: DFT-based MD and model systems
DFT-based Born–Oppenheimer molecular dynamics (BOMD) simulations were performed using the Becke ex-change [36] and Lee–Yang–Parr [37] correlation functionals.All the calculations have been carried out with the freelyavailable DFT package CP2K/Quickstep [38], which isbased on the hybrid Gaussian and plane wave method [39].Analytic Goedecker–Teter–Hutter [40, 41] pseudopotentials,a TZV2P level basis set for the orbitals and a density cutoffof 280 Ryd were used. Only the 0-point was considered.We study the neutral surfaces of (0001) ↵-quartz and (0001)↵-alumina, corresponding to low pH (
X-ray absorption Fine Structure
15
Characterization of materials by synchrotron radiation
X-ray spectra connect to the core-electron excited states (role of core hole)
Element specific probe:
electronic structure in situ
time resolution (fs)
no long range order required
imaging
Challenging interpretation of the spectra
Central role of theoretical approach
charge transfer
nature of bonding
hybridization
chemical environment
X-ray Absorption Simulations
16
Interpretation of experiment, structures refinement (signal assignment),
understanding of physical-chemical properties of materials.
The theoretical spectra of Gly are shown below the experi-
mental spectra in Figure 1b,c as bar diagrams with bar heights
proportional to the oscillator strength. These computed spectra
are discussed later in conjunction with the results of the STEX
calculations. Here we note that in both standard and polarized
ion STEX approximations, the main peak at ∼289.5 eVoriginates from the C 1s(CR)fσ*CNH excitations. As anindependent confirmation we have measured NEXAFS spectra
of Gly thin films crystallized from aqueous solutions having
different pH, and the specimens with pH > pK2 (above whichthe amino group loses a proton) show different NEXAFS
features at ∼290 eV.9 This is clearly a deviation from the simplebuilding block approach since the NEXAFS spectrum cannot
be reduced to the sum of pair bond features; instead, the whole
amino group (more accurately -CNH3) needs to be consideredas a building block. To summarize:
• First, although the sample preparation procedure leaves theamino acid molecules in a charged state, the low-energy part
of the spectra does not exhibit any additional structure with
respect to the zwitterionic state and so can be directly compared
with our calculations of the C K edge spectrum of the
zwitterionic form.
• Second, we can identify two building blocks common forall amino acids: the carboxyl -COOH group, and the -CNH3group. The carboxyl group leads to the sharp, intense line at
288.65 eV, assigned to the C 1s-1π* state. The σ* excitationsoriginating from the C-O bond of the carboxyl group appearon the high-energy side of the spectra and can be observed either
as two separate peaks (protonated), or one broad (deprotonated)
peak depending on the charge state.
• Third, the methyl group, which leads to a low-energy peakat ∼288 eV in chemisorbed Gly,26,27 cannot be considered as agood building block. Instead the -CNH3 group, which can beidentified by the peaks at ∼290.8 eV (with mostly C 1s-1σC-N*character) and the shoulder at 289.3 eV, can be used as a
building block.
3.2. Amino Acids with Simple Aliphatic Side Chains:
Alanine, Valine, Leucine, Isoleucine. Aliphatic amino acids,
such as alanine (Ala), valine (Val), leucine (Leu), and isoleucine
(Ile), have nonpolar side chains which do not bind to a surface,
which do not release protons, and which do not participate in
hydrogen or ionic bonds. The side chains of these amino acids
can be thought of as “oily” or lipid-like, a property that promotes
hydrophobic interactions. NEXAFS spectra of this class of
amino acids are shown in Figure 2 while the energies, term
values, and proposed assignments of spectral features are
summarized in Table 2. The spectra were normalized to atomic
scattering factors30 in order to give a direct measure of the amino
acid mass absorption strength. The lower energy side of each
spectrum was fit with a combination of Gaussian functions in
order to determine the spectroscopic trend. The separate peak
contributions are drawn in thin lines and the fitted spectrum is
shown as a continuous line drawn through the experimental
points shown by markers. We have restricted the analysis to
the part below 292 eV in order to avoid overlap with the TFA
peaks. The high-energy side is fit by a step function, broadened
by a Gaussian. The number of Gaussian lines used, as well as
their parameters, were determined from optimization of a free
fit, i.e., without any parameter constraints. This likely leads to
a situation where contributions of individual states remained
unresolved in some cases and so appear as a single broad band.
In particular, as in the case of Gly, both σ*C-C and σ*C-N bondscontribute to the oscillator strength near the IP. In the present
analysis their contributions cannot be distinguished from near-
IP excitations, so the energy position of the step function, as
well as its Gaussian broadening, are left to be determined by
the fitting procedure, and thus the step function represents all
near-IP excitations.
The fit reveals four groups of peaks present in each C K edge
spectrum, as shown in Figure 2. All experimental spectra are
dominated by a strong peak at 288.65 ( 0.05 eV that can beassigned unambiguously to C 1s(COO)f π* transitions. Exceptfor a small shift from Gly to Ala (
17
XES/NEXAFS local probes for electronic and geometric properties
Occupied orbitals
Two step process Final state : valence-hole
Binding of adsorbed molecules
Unoccupied levels, symmetry resolved
One step process Finale state: core-hole
Instantaneous configuration in dynamic systems
3.1 C ore L evel Sp ec t roscop y 37
NON-RADIANTAuger Electron Spectroscopy (AES)
RADIANTX-ray Emission Spectroscopy (XES)
hν
Figure 3.5: Schematic illustration of non-resonant core-hole decay processes
decay, schematically described in Figure 3.5 for an initial state prepared by core-holeionization. In the radiant decay, an electron in the occupied valence band fills the corehole, and a photon is emitted, forming the basis for x-ray emission spectroscopy (XES). Thenon-radiant decay mechanism is very different. In this transition, no photon is involvedand it is therefore not subject to the electric dipole interaction picture. The final statetypically consists of two holes in the valence shell and an ejected electron, the Augerelectron. Measurements of the ejected electron energy distribution forms the basis forAuger Electron Spectroscopy (AES).
The cross-sections for radiative and non-radiative decay of a core-hole state are stronglyelement dependent. For the low Z elements carbon, nitrogen and oxygen, the Auger rateτa is much faster than the fluorescence rate τf ; the probability of radiative decay is 0.52%,0.83% and 1.3%, respectively 115. The total lifetime of the core-hole τh, given by thedifferent decay rates according to 1τ =
1τh
+ 1τe , can be determined by measuring thelifetime broadening in the core creation processes (see previous sections). It can be shownthat τh is rather insensitive to the chemical state of the atom and thus considered aninherent element property. The lifetime of the core-hole is not known exactly, and in thecase of oxygen, experimental determination of the lifetime broadening varies in the energyrange 0.165 ± 0.2 eV (corresponding to ∼ 4 fs ) 76 ,107. In this study, the value 3.6 fs forthe core-hole lifetime is used 76.
In the normal treatment of AES and XES one usually uses a two-step language, i.e.the core-hole creation (Fig. 3.1) and the core-hole decay (Fig. 3.5) processes are consec-utive and independent events. However, a number of high-resolution studies have clearlydemonstrated that for resonant excitations, the two-step picture is not generally valid(see 116 and references therein). Rather a one-step treatment is required.
Consider the radiative decay following core-hole excitation. In the one-step formalismthe excitation and decay of the core-excited state are treated as one inseparable scatteringevent. The intermediate final states |m⟩ are considered (virtual) resonances with a lifetimebroadening Γm. This is captured theoretically in second-order perturbation theory of thephoton-electron interaction, leading to the Kramers-Heisenberg scattering formula thatwas introduced in section 2.3. In particular, this description contains the sum of matrixcross terms proportional to
⟨f | e′ · p |m⟩ ⟨m| p · e | i⟩ (3.7)
These unsquared terms cannot be easily visualized as probabilities, and it is not correct toassume that a specific intermediate state is excited with a certain probability in the scat-tering process. This is the crucial difference between the one- and the two-step pictures.In the one step formalism described above, the sum over intermediate states |m⟩ leads to
core states
bounded states
continuum states
IP
core states
bounded states
continuum states
IP
hν
NEXAFS
XES
30 Methods
hν
hν
IONIZATIONX-ray Photoelectron Spectroscopy (XPS)
EXCITATIONX-ray Adsorption Spectroscopy (XAS)
IP
bound states
continuum states
Core Levels
Figure 3.1: Schematic illustration of core-hole creation processes
following, we will discuss in more detail what we can learn from these spectroscopies. Wedivide the core level spectroscopies into two classes; core-hole creation (XPS, XAS) andcore-hole decay (XES, AES). A one-electron orbital picture is often adopted in line withsection 2.2.2, and both the Born-Oppenheimer (section 2.2.1) and dipole approximation(section 2.3.1) are applied throughout this section. The photon source is considered 100%linearly polarized which is a good approximation for undulator radiation (section 3.3.1).Results from the present work (if not otherwise stated) serve as examples for the introducedconcepts, and some aspects that are not directly related to this thesis have been omitted.
3.1.1 Core-Hole Creation
The absorption of a photon can result in ionization (XPS) or excitation (XAS) of a coreelectron depending on the energy h̄ω of the photon. For a detailed description of theseprocesses consider the textbooks by Hüffner104 and Stöhr105. Schematic diagrams ofXPS and XAS are shown in Figure 3.1, where the orbitals of the free water moleculeare used to represent an arbitrary N electron system. In the x-ray absorption process theincoming X-ray photon is absorbed and an electron from an inner shell is transferred into anunoccupied state in the valence band. The energy distribution of the excitation probabilityinto unoccupied orbitals forms the XA spectrum. If the energy of the photon h̄ω islarger than the core electron ionization potential (IP), the electron is removed completelyfrom the molecule (transferred to a continuum of states given by a free electron in thepresence of a core-hole potential). Here, the distribution of ionized core electrons forms thephotoelectron spectrum. Obviously, the valence electrons can also be emitted, normallydenoted UPS (UV photoelectron spectroscopy) or just PES (photoemission spectroscopy).The energy conservation law implies that the transition energy h̄ω is given as a differencein total energy before and after the spectroscopic event
h̄ω = Ef − Ei (3.1)
where Ei,f is the total energy before and after the core excitation (ionization). In thecreation of the core-hole an attractive coulombic potential at the ionized site is created.The system will of course respond to this potential to find the lowest energy configurationin the presence of the core hole. For free molecules this generally means contraction oforbitals at the core-hole site, whereas neighboring molecules are polarized. Charge-transferscreening is also possible, for metallic systems this is denoted metallic screening.
In the ionization event, the ionized electron leaves a system on a time-scale dependenton the kinetic energy. For photon energies well above the ionization potential (∼ 10 eV
30 Methods
hν
hν
IONIZATIONX-ray Photoelectron Spectroscopy (XPS)
EXCITATIONX-ray Adsorption Spectroscopy (XAS)
IP
bound states
continuum states
Core Levels
Figure 3.1: Schematic illustration of core-hole creation processes
following, we will discuss in more detail what we can learn from these spectroscopies. Wedivide the core level spectroscopies into two classes; core-hole creation (XPS, XAS) andcore-hole decay (XES, AES). A one-electron orbital picture is often adopted in line withsection 2.2.2, and both the Born-Oppenheimer (section 2.2.1) and dipole approximation(section 2.3.1) are applied throughout this section. The photon source is considered 100%linearly polarized which is a good approximation for undulator radiation (section 3.3.1).Results from the present work (if not otherwise stated) serve as examples for the introducedconcepts, and some aspects that are not directly related to this thesis have been omitted.
3.1.1 Core-Hole Creation
The absorption of a photon can result in ionization (XPS) or excitation (XAS) of a coreelectron depending on the energy h̄ω of the photon. For a detailed description of theseprocesses consider the textbooks by Hüffner104 and Stöhr105. Schematic diagrams ofXPS and XAS are shown in Figure 3.1, where the orbitals of the free water moleculeare used to represent an arbitrary N electron system. In the x-ray absorption process theincoming X-ray photon is absorbed and an electron from an inner shell is transferred into anunoccupied state in the valence band. The energy distribution of the excitation probabilityinto unoccupied orbitals forms the XA spectrum. If the energy of the photon h̄ω islarger than the core electron ionization potential (IP), the electron is removed completelyfrom the molecule (transferred to a continuum of states given by a free electron in thepresence of a core-hole potential). Here, the distribution of ionized core electrons forms thephotoelectron spectrum. Obviously, the valence electrons can also be emitted, normallydenoted UPS (UV photoelectron spectroscopy) or just PES (photoemission spectroscopy).The energy conservation law implies that the transition energy h̄ω is given as a differencein total energy before and after the spectroscopic event
h̄ω = Ef − Ei (3.1)
where Ei,f is the total energy before and after the core excitation (ionization). In thecreation of the core-hole an attractive coulombic potential at the ionized site is created.The system will of course respond to this potential to find the lowest energy configurationin the presence of the core hole. For free molecules this generally means contraction oforbitals at the core-hole site, whereas neighboring molecules are polarized. Charge-transferscreening is also possible, for metallic systems this is denoted metallic screening.
In the ionization event, the ionized electron leaves a system on a time-scale dependenton the kinetic energy. For photon energies well above the ionization potential (∼ 10 eV
Core-hole Creation and Decay
Interaction with an electromagnetic field (incoming radiation)
Linear perturbation term (time dependent perturbations)
Induced transition: probability according to Fermi golden rule
Long wavelength (UV, soft x-rays)
Equivalent operators
Interaction X-ray with matter
18
A(r, t) = A0 e cos (k · r − ωt)
V (t) =e
mcA · p
Pif =πe2
2h̄m2c2A20|⟨f |e
ik·re · p|i⟩|2ρi(E)
krs =2π
λrs ≪ 1
p̂ = �i�⇥ = im� [Ĥ, r̂]
velocity form
Pif ⇥ (Ef � Ei) |⇤f | µ̂ |i⌅|2
Pif � |⇥f | ê · p̂ |i⇤|2
length form
∆ SCF
Ground state KS virtual orbitals
Parametric SCF solution
Numerically not stable : difference between 2 SCF calculations
Higher excited states states (collapse)
Not orthogonal MOs sets
19
IPi ̸= −εi ∆E(1s → π∗) ̸= επ∗ − ε1s
n(r) =∑
i
fi |ψi(r)|2
Ẽ[n({fi})]
IPi=1 = Ẽ(01, 12.., 1N , ...0r...) − Ẽ(11, 12.., 1N , ..., 0r, ...)
∆E(i → r) = Ẽ(11, 12..., 0i, ..., 1N , ..., 1r, ...) − Ẽ(11, 12..., 1i, ...1N ..., 0r, ...)
Pif ∝ |⟨f |e · p|i⟩|2ρi(E)
The Transition Potential Method
Direct calculation of the excitation energies from the solution of KS equation with a modified core potential on the absorbing atom
Relaxation effects up to second order in ∂E/∂f : balance between initial and final state contributions
One calculation for each excitation
20
IPi =
∫ 0
1
dfi∂Ẽ({fj})
∂fi≃
(
∂Ẽ({fj})
∂fi
)
fi=1
2
= εi
(
1
2
)
h̄ωif = εTf − ε
Ti Iif =
2
3ωif |⟨ψ
Ti |µ̄|ψ
Tf ⟩|
2
Independent-Particle Approach
Errors/Corrections
Single particle picture (hole/electron correlation)
Basis sets flexible to relaxation to describe Rydberg states; Double basis set approach [19s,19p,19d]
Using TP: error due to higher order contributions to the core relaxation energy
Energy corrections based on ∆ SCF
Relativistic correction as rigid, species-dependent translation
21
∆ IP = εT1s
(
1
2
)
− IP∆SCF
∆ (1s → π∗) =
(
επ∗ − εT1s
(
1
2
))
− ∆E∆SCF
The Core Hole
22
The actual location of the excited electron in the final state has a relative small effect on relaxation of the orbitals
The focus is set on the core hole
Only one SCF calculation per core hole for the entire spectrum
HCH half core hole
system charged +1/2
contribution from the
initial state
XHCH half core hole
half 1st excited state neutral system
contribution from the
initial state
FCH full core hole
system charged +1
dominated by the
final state
XFCH full core hole
full 1st excited state neutral system
contribution by the
final state
AE-TP in Condensed Matter
Ground state SCF and localization of the occupied orbitals
Character of the Core states
Half Core Hole or Full Core Hole (LSD)
SCF with modified occupation numbers
Oscillator strengths in the velocity form
23
ψ̃i(r) =∑
µ
C̃µiϕµ(r) Maxi
{〈
ψ̃Oni |ϕSTOO,1s
〉}
⇒O1s, n
⇒ j-th state
fj =1
2, 0
Iif = |⟨ψTi |∇|ψ
Tf ⟩|
2
Input XAS
24
&XAS &SCF EPS_SCF 1.0E-4 MAX_SCF 200 &SMEAR ON METHOD FERMI_DIRAC ELECTRONIC_TEMPERATURE [K] 300 &END &MIXING METHOD BROYDEN_MIXING ALPHA 0.1 BETA 1.5 NBUFFER 8 &END MIXING &END SCF METHOD TP_HH DIPOLE_FORM VELOCITY STATE_TYPE 1s STATE_SEARCH 1 ATOMS_LIST 1 ADDED_MOS 1000 &LOCALIZE &END
&PRINT &PROGRAM_RUN_INFO &END &RESTART FILENAME ./hch_dim_p12_ud &EACH XAS_SCF 15 &END ADD_LAST NUMERIC &END &XAS_SPECTRUM FILENAME ./hch_dim_p12_ud &END &XES_SPECTRUM FILENAME ./hch_dim_p12_ud &END &END &END XAS
Broadening and Shift
25
32 3 Methodology of NEXAFS calculations
with typical values as:
⇥min ⇥ 0.5 eV, ⇥max ⇥ 8 eV, Emax � Emin ⇥ 20 eV (3.8)
Figure 3.2: Example of a calculation: the spikes resulting from a TPcalculation are convoluted with a superposition of Gaussians as described inEq. (3.7) (red). The transitions are shifted to the �SCF value of the firsttransition (dotted line) by matching this value with the first TP resonance(⇥ 290 eV) by an alignment of �� and the final spectrum is obtained (green).
Convolution using Gaussian functions with adapted width
Energy (eV)
26
1
Energy [eV] Energy [eV] Energy [eV]295 305285 295 305285
Exp. Exp.
ortho
meta
paraC1sC1s N1s
398 406402 410
Pyridineπ*
M P
O
Small Molecules: atomic environment
Para
Meta
Ortho
Iannuzzi, Hutter, PCCP (2007)
Ortho Meta Paraπ* π* π*
3π*+Ry
Chemical Shift
π*
σ*CH3π*+Ryσ*CH
27
Energy (eV)310290 300
Exp.
Diamond
Extended Systems: supercell approach
512 C atoms pristine crystal
6-311G(dp) PBE
Core 1s exciton
FIG. 3: Snapshot of the final compressed state. The light grey spheres denote sp3 hybridized
carbon centers and the dark grey spheres denote sp2 hybridized centers.
cubic diamond is in good agreement with that obtained from the final shock compressed
state.
In conclusion, we have presented the first ab initio calculations of graphite under uniaxial
shock compression using state-of-the-art software and hardware. Our results indicate that
the transition from graphite to a diamond-like state is Martensitic, in agreement with exper-
imental conjectures [3].We observe herein an intermediate layered diamond phase is formed
without plane sliding, a completely different transformation mechanism than is found under
hydrostatic conditions. Because of the time-scales of our simulation it is unlikely that we
are capturing the final hydrostatic diamond-like state, thus the use of the MSSM method
and its ability to capture physical intermediate states is crucial. From Figs. 4 and 2 it is
evident that all of the features of a diamond-like state are present and the XAS is in good
agreement with DAC studies [2].
This work was performed under the auspices of the U.S. Department of Energy by Uni-
versity of California LLNL under contract No. W-7405-Eng-48. We would like to thank
Tony Van Buuren, Trevor Willey, Dave Erskine, and Jon Eggert for illuminating discus-
sions. We also gratefully acknowledge Mike McCoy and Dale Nielsen of the Center for
Applied Scientific Computing.
7
FIG. 3: Snapshot of the final compressed state. The light grey spheres denote sp3 hybridized
carbon centers and the dark grey spheres denote sp2 hybridized centers.
cubic diamond is in good agreement with that obtained from the final shock compressed
state.
In conclusion, we have presented the first ab initio calculations of graphite under uniaxial
shock compression using state-of-the-art software and hardware. Our results indicate that
the transition from graphite to a diamond-like state is Martensitic, in agreement with exper-
imental conjectures [3].We observe herein an intermediate layered diamond phase is formed
without plane sliding, a completely different transformation mechanism than is found under
hydrostatic conditions. Because of the time-scales of our simulation it is unlikely that we
are capturing the final hydrostatic diamond-like state, thus the use of the MSSM method
and its ability to capture physical intermediate states is crucial. From Figs. 4 and 2 it is
evident that all of the features of a diamond-like state are present and the XAS is in good
agreement with DAC studies [2].
This work was performed under the auspices of the U.S. Department of Energy by Uni-
versity of California LLNL under contract No. W-7405-Eng-48. We would like to thank
Tony Van Buuren, Trevor Willey, Dave Erskine, and Jon Eggert for illuminating discus-
sions. We also gratefully acknowledge Mike McCoy and Dale Nielsen of the Center for
Applied Scientific Computing.
7
Graphite under shock compression
uniaxial loading shock 12 km/s (>130GPa)
Martensitic transition graphite->diamond
high defect concentration
sp3 C
sp2 C
π*
No
rmalized
In
ten
sit
y
280 290 300 310 320Energy [eV]
FIG. 4: Calculated XAS of (top panel) graphite at 300K, (middle panel) our final compressed
state, (bottom panel) hydrostatically compressed cubic diamond at the same density as the final
compressed state. The XAS of graphite shows the π∗ pre-edge intensity at ∼ 285 eV and the
σ∗ intensity at ∼ 295 eV. The middle panel shows the XAS spectrum averaged over 200 carbon
centers. The gray curve in the middle panel is a normalized partial XAS spectra of only the sp2
carbon centers.
[1] S. Naka, K. Horii, Y. Takeda, and H. Hanawa, Nature 259, 38 (1976).
[2] W. L. Mao, H. K. Mao, P. J. Eng, P. T. Trainor, M. Newville, C. C. Kao, D. L. Heinz, J. Shu,
Y. Meng, and R. J. Hemley, Science 302, 425 (2003).
[3] D. J. Erskine and W. J. Nellis, Nature 349, 317 (1991); D. J. Erskine and W. J. Nellis, J.
Appl. Phys. 71, 4882 (1992).
[4] T. Irifune, A. Kurio, S. Sakamoto, T. Inoue, and H. Sumiya, Nature 421, 599 (2003).
[5] S. Scandolo, M. Bernasconi, G. L. Chiarotti, P. Focher, and E. Tosatti, Phys. Rev. Lett. 74,
4015 (1995).
[6] F. J. Ribeiro, P. Tangney, S. G. Louie, and M. L. Cohen, Phys. Rev. B 72, 214109 (2005).
[7] A. Sorkin, J. Alder, and R. Kalish, Phys. Rev. Lett. 74, 064115 (2006).
[8] R. J. Hemley, Annual Review of Physical Chemistry 51, 763 (2000).
[9] J. D. Kress, S. R. Bickham, L. A. Collins, B. L. Holian, and S. Goedecker, Phys. Rev. Lett.
83, 3896 (1999).
8
Energy (eV)280 320300
second gap
Compressed Diamond
Graphite
Final State
σ*
Mundy, Curioni, Kuo, Iannuzzi, et al, JCP, 128, (2008)
Magnetic Shielding
28
Interaction between B and a spin 1/2 nucleus with spin angular momentum
in diamagnetic materials the Bind arises only from the induced orbital currents j
second derivative of electronic energy
chemical shift w.r.t. reference
Schielding and g tensors
29
applied B0 along x
gxy
= gZxy
+�gZKExy
+�gSOxy
+�gSOOxy
⇢s(r) = ⇢↵(r)� ⇢�(r) spin density
gZxy
= ge
�xy
�gZKExy
= �gec2
�T↵ � T �
��xy
�gSOxy
=ge
� 1c
Z
⌦C
hj↵x
⇥rV ↵e↵ � j�x ⇥rV�
e↵
id3r
�gSOOxy
= 2
Z
⌦C
Bcorrxy
(r)⇢s(r)d3r
free el. g value
unperturbed kinetic energy
induced spin current effective pot.
magnetic field from current
Linear Response
The magnetic field B represented by the vector potential A
First order and second order perturbation terms in the Hamiltonian
First order correction of the electronic ground state
imaginary response n1=0
�iX
il
✓Hkl�ij � Skl
Z (0)i (r)H(r)
(0)j (r)dr
3
◆C(1)ij =
X
l
H(1)kl(j)C0lj
Gauge origin problem
31
Convergence w.r.t basis set strongly depend on gauge Rg=d(r)
Gauge invariance in molecular and periodic systems
GIAO accurate but involved and not efficient
IGLO
IGAIM gauge terms on the perturbation functional
CSGT
no diamagnetic term, computationally convenient, slow converging with basis set quality
Induced Current
diamagneticparamagnetic
Calculation of Induced Currents
32Sebastiani, Parrinello, JPC, 105, 1951 (2001)
Partitioning of the perturbation operator
dependency on ψ0
x-component of linear current density j induced from B0 applied along y
DFPT
diamagnetic
GAPW Chemical Shift
33Weber, Iannuzzi, Giani, Hutter, Declerck, Waroquier, JCP, 131, 014106 (2009)
GAPW induced
current density
applied B0 along x
from soft term in reciprocal space
from local terms by integration on spherical grids
NMR Input
34
&GLOBAL PROJECT ${PROJECT_NAME} PRINT_LEVEL LOW RUN_TYPE LINEAR_RESPONSE&END GLOBAL
Linear Response run-type
&DFT BASIS_SET_FILE_NAME EMSL_BASIS_SETS POTENTIAL_FILE_NAME POTENTIAL &MGRID CUTOFF 300 &END MGRID &QS METHOD GAPW &END QS &SCF &OUTER_SCF MAX_SCF 200 &END OUTER_SCF MAX_SCF 20 EPS_SCF 10E-6 &OT ALGORITHM IRAC PRECONDITIONER FULL_ALL &END OT &END SCF &XC &XC_FUNCTIONAL BLYP &END XC_FUNCTIONAL &END XC &END DFT
Ground State GAPW
&PROPERTIES &LINRES &LOCALIZE &END OPT_METHOD CG LINESEARCH 2PNT PRECONDITIONER FULL_ALL MAX_SCF 2000 &CURRENT GAUGE ATOM ORBITAL_CENTER COMMON &END CURRENT &NMR INTERPOLATE_SHIFT T &PRINT &CHI_TENSOR &END CHI_TENSOR &END PRINT &END &END &END
NMR Chemical Shift
&KIND N LEBEDEV_GRID 100 RADIAL_GRID 200 BASIS_SET aug-cc-pVQZ POTENTIAL ALL &END KIND
GAPW kind
NMR Output for SiH4
35
Total electronic density (r-space): -11.6551998133 6.3448001867 Total core charge density (r-space): 18.0000000000 0.0000000000
Hard and soft densities (Lebedev): -67.3124445208 -60.9676446784 Total Rho_soft + Rho1_hard - Rho1_soft (r-space): -17.9999996557 Total charge density (r-space): 0.0000003443 Total Rho_soft + Rho0_soft (g-space): 0.0000003075
Overlap energy of the core charge distribution: 0.00000011310807 Self energy of the core charge distribution: -131.30230308020512 Core Hamiltonian energy: -159.90960164160288 Hartree energy: 32.53453281991028 Exchange-correlation energy: -5.66831828118873
GAPW| Exc from hard and soft atomic rho1: -16.66150292615432 GAPW| local Eh = 1 center integrals: -10.87514648571372
outer SCF iter = 2 RMS gradient = 0.35E-05 energy = -291.8823394818 outer SCF loop converged in 2 iterations or 23 steps
ENERGY| Total FORCE_EVAL ( QS ) energy (a.u.): -291.882339481846373
Converged GAPW SCF
NMR Output for SiH4
36
LOCALIZE| The spread relative to a set of orbitals is computed LOCALIZE| Orbitals to be localized: All the occupied LOCALIZE| Spread defined by the Berry phase operator LOCALIZE| The optimal unitary transformation is generated by the Jacobi algorithm
Localization of the ground state orbitals before starting the linear response calculation Localization by iterative Jacobi rotation Iteration Tolerance 100 0.2560E-01 200 0.1556E-01 300 0.6506E-02 400 0.1981E-02 500 0.8639E-03 600 0.3288E-03 Localization for spin 1 converged in 698 iterations Total Spread (Berry) x,y,z: 1.1135942650 1.1135229959 1.1133052793
WANNIER CENTERS for spin 1
--------------- Centers --------------- --------------- Spreads --------------- state 1 -0.222895 0.146450 0.058049 0.541470 0.000000 0.000000 state 2 0.000107 -0.064397 -0.264849 0.541498 0.000000 0.000000 state 3 0.222716 0.146673 0.058175 0.541470 0.000000 0.000000 state 4 0.000000 0.000121 0.000245 0.140158 0.000000 0.000000 state 5 0.000072 -0.228327 0.149401 0.541477 0.000000 0.000000 state 6 -1.772371 -1.183928 -0.406835 1.888384 0.000000 0.000000 state 7 0.000093 0.608045 2.082470 1.887723 0.000000 0.000000 state 8 0.000211 1.760315 -1.267771 1.887975 0.000000 0.000000 state 9 1.772063 -1.184376 -0.406863 1.888382 0.000000 0.000000
Localization procedure through Jacobi rotations
NMR Output for SiH4
37
============================================================================= START LINRES CALCULATION =============================================================================
Properties to be Calulated:
NMR Chemical Shift LINRES| LOCALIZED PSI0 LINRES| Optimization algorithm Conjugate Gradients LINRES| Line Search 2pnt LINRES| Preconditioner FULL ALL
LINRES| EPS_SCF 1.0E-05
LINRES| Max number of iterations per SCF cycle 2000
Total energy ground state: -291.88233948447117 -65.49055205 -5.05920132 -3.47810739 -3.47810596 -3.47809942 -0.48735526 -0.30828056 -0.30810235 -0.30804975
*** Start current Calculation ***
Initialization of the current environment
CURRENT| Gauge used ATOM CURRENT| Orbital center used COMMON CURRENT| Common center 0.000000 0.000000 0.000000 CURRENT| Calculation of the p and (r-d)xp operators applied to psi0
Linear Response Initialization
NMR Output for SiH4
38
*** Self consistent optimization of the response wavefunctions ***
Response to the perturbation operator P_x
Iter. Method Stepsize G-norm Convergence Total energy ----------------------------------------------------------------------------- 10 LR LS 0.99E-01 0.0000000000 0.0323894903 -2.2638501245 20 LR LS 0.12E+00 0.0000000000 0.0077609426 -2.6500765218 30 LR LS 0.48E-01 0.0000000000 0.0077044886 -2.6794773033 40 LR LS 0.15E+00 0.0000000000 0.0050879317 -2.7024052890 50 LR LS 0.60E+00 0.0000000000 0.0074222310 -2.7361724629 60 LR LS 0.55E-01 0.0000000000 0.0036544263 -2.7922066072 70 LR LS 0.68E-01 0.0000000000 0.0027598118 -2.7965811305 80 LR LS 0.90E-01 0.0000000000 0.0011411853 -2.7983145616 90 LR LS 0.11E+00 0.0000000000 0.0008189226 -2.7988545411 100 LR LS 0.72E-01 0.0000000000 0.0005977973 -2.7991253972 110 LR LS 0.73E-01 0.0000000000 0.0003155523 -2.7991931045 120 LR LS 0.64E-01 0.0000000000 0.0002261754 -2.7992267386 130 LR LS 0.76E-01 0.0000000000 0.0000928637 -2.7992365112 140 LR LS 0.62E-01 0.0000000000 0.0000783777 -2.7992406728 150 LR LS 0.94E-01 0.0000000000 0.0000318069 -2.7992417209 160 LR LS 0.90E-01 0.0000000000 0.0000245090 -2.7992422317 The SCF has converged in 167 iterations
Second order energy P_x= -0.2799242338698460E+01 Store the psi1 for the calculation of the response current density Write the resulting psi1 in restart file... not implemented yet
Localization procedure through Jacobi rotations
NMR Output for SiH4
39
Initialization of the NMR environment
NMR| Shift factor (ppm) 9.863793E-04 NMR| Shift factor gapw (ppm) 5.325134E+01 NMR| Chi factor (SI) 1.972757E+01 NMR| Conversion Chi (ppm/cgs) 6.022045E-02 NMR| Conversion Chi to Shift 1.047198E-04 NMR| Shielding tensor computed for 5 atoms Integrated j_xx(r): G-space= 0.4608768062028217E+00 R-space= 0.4608768062028132E+00 Integrated j_yx(r): G-space= -0.2077039113301925E+01 R-space= -0.2077039113301958E+01 Integrated j_zx(r): G-space= -0.2039298825792812E+01 R-space= -0.2039298825793122E+01 CheckSum R-integrated j= 0.2947072888403525E+01 WARNING WARNING WARNING WARNING WARNING WARNING WARNING chi_analytic: gapw= T WARNING WARNING WARNING WARNING WARNING WARNING WARNING Integrated local j_xx(r)= -0.9615000648009942E-06 Integrated local j_yx(r)= -0.4802383200079088E-04 Integrated local j_zx(r)= 0.3935076278575146E-04
Final NMR output
CheckSum Chi = 0.6343893866003933E+03 CheckSum Shifts = 0.8086834144167068E+03
============================================================================= ENDED LINRES CALCULATION =============================================================================
.......
IGAIM-GAPW
40
Test set composed of 26 small molecules C2H2 , CH2O, CH3Cl, CH3F, CH4, Cl2, CO2, F2, FCl, H2O2, H2O, H2, H2S, HCl, HCN, HCOOH, HF, HNO3, N2O, N2, NH3, NO2, O2, O3, PH3, SO2
the Gaussian representation of the current density is also farfrom trivial. However, it generally holds that !gxy
SOO is arelatively small term in comparison with !gxy
ZKE and !gxySO.19
For this reason, we choose to neglect the contributions fromthe atom-centered current densities jxA
corr− j̃xAcorr to the G!0
components B̃x,G!0corr of the induced magnetic field. Being
computed from the soft current density j̃xcorr only, on the
reciprocal-space FFT-grid, these components will be soft too!hence the tilde". Bx,G=0
corr , on the other hand, will be computedanalytically via Eq. !9", using the Gaussian representation ofthe current density. In summary, !gxy
SOO is evaluated as fol-lows:
!gxySOO # 2$
FFTB̃xy,G!0
corr !r""̃s!r"d3r + 2%A$
UA
B̃xy,G!0corr !r"
#!"As !r" − "̃A
s !r""d3r + 2Bxy,G=0corr .
III. RESULTS AND DISCUSSION
A. Test calculations
All developments were implemented in QUICKSTEP,which is part of CP2K.36 CP2K is a freely available under theGNU General Public License !GPL" general program to per-form atomistic and molecular simulations of solid state, liq-uid, molecular, and biological systems. A description ofQUICKSTEP can be found in Ref. 37.
The GAPW method for the AE calculation of the NMRchemical shifts and the EPR g tensor was validated by com-parison, for a series of small isolated molecules, with theresults from conceptually similar gas-phase methods, !i" theNMR routines as implemented in the G03 program package38
!further referred to as the IGAIM-GA method", and !ii" the gtensor method of Schreckenbach and Ziegler34 as imple-mented in the ADF !Ref. 39" program package !further re-ferred to as the SZ method". Isolated molecules are approxi-mated by a supercell approach with a large cell size of!20 Å"3. We used a 300 Ry cutoff for the auxiliary planewave grid, the BLYP40,41 gradient-corrected exchange-correlation functional, and the efficient and numericallystable orbital transformation energy minimizer introduced inRef. 42. The Gaussian basis sets used in this work weretaken from the Environmental Molecular Sciences Labora-tory !EMSL" basis set exchange library.43
1. NMR
A comparison between the IGAIM method implementedin G03 !IGAIM-GA" and the presented IGAIM-GAPWmethod for the isotropic and the anisotropic chemical shiftsof a representative set of small molecules is shown in Fig. 1.Three different basis sets were used, namely, the Pople split-valence double-zeta plus polarization 6-31G!d , p",44,45 andthe augmented double and triple-zeta correlation consistentbasis sets aug-cc-pVDZ and aug-cc-pVTZ, respectively46–49.The test set is composed of 26 molecules, namely: C2H2,CH2O, CH3Cl, CH3F, CH4, Cl2, CO2, F2, FCl, H2O2, H2O,H2, H2S, HCl, HCN, HCOOH, HF, HNO3, N2O, N2, NH3,
NO2, O2, O3, PH3, and SO2. All the geometries were opti-mized at the BLYP/aug-cc-pVDZ level of theory.
For the small 6-31G!d , p" basis set, a maximal unsignedrelative error of 30% is observed for the isotropic chemicalshift of the carbon atom in HCN. The isotropic shift for thenitrogen atom in HCN is 3.9 ppm for IGAIM-GA and 5.1ppm for IGAIM-GAPW. This absolute error !1.2 ppm" is ofthe same order as for other nitrogen atoms with, e.g., 0.6ppm for NH3. Except for this extreme case, the other isotro-pic !anisotropic" chemical shifts have a smaller maximal un-signed relative error with, e.g., less than 1% !10%" for hy-drogen, 5% !0.5%" for carbon, and 2% !2%" for oxygen.
The unsigned relative error is greatly reduced for thelarger aug-cc-pVDZ and aug-cc-pVTZ basis sets. Theaug-cc-pVDZ provides a maximal unsigned relative error of5% !nitrogen in HCN" and 10% !H2" for, respectively, theisotropic and anisotropic components of the shift. The maxi-mal unsigned relative errors for the isotropic shift are 0.5%
-30
-20
-10
0
10
20
30
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/6-31G**
-12
-6
0
6
12
rela
tive
erro
rin
%isotropicanisotropic
BLYP/aug-cc-pVDZ
0 20 40 60 80test set
-4
-2
0
2
4
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/aug-cc-pVTZ
FIG. 1. Relative error of the IGAIM-GAPW method with respect to theIGAIM-GA method for the isotropic and anisotropic chemical shifts calcu-lated on a set of small molecules !see text for details". Note the differentscales in the three panels.
014106-5 Magnetic linear response properties J. Chem. Phys. 131, 014106 "2009#
Downloaded 27 Jan 2011 to 130.60.136.86. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
the Gaussian representation of the current density is also farfrom trivial. However, it generally holds that !gxy
SOO is arelatively small term in comparison with !gxy
ZKE and !gxySO.19
For this reason, we choose to neglect the contributions fromthe atom-centered current densities jxA
corr− j̃xAcorr to the G!0
components B̃x,G!0corr of the induced magnetic field. Being
computed from the soft current density j̃xcorr only, on the
reciprocal-space FFT-grid, these components will be soft too!hence the tilde". Bx,G=0
corr , on the other hand, will be computedanalytically via Eq. !9", using the Gaussian representation ofthe current density. In summary, !gxy
SOO is evaluated as fol-lows:
!gxySOO # 2$
FFTB̃xy,G!0
corr !r""̃s!r"d3r + 2%A$
UA
B̃xy,G!0corr !r"
#!"As !r" − "̃A
s !r""d3r + 2Bxy,G=0corr .
III. RESULTS AND DISCUSSION
A. Test calculations
All developments were implemented in QUICKSTEP,which is part of CP2K.36 CP2K is a freely available under theGNU General Public License !GPL" general program to per-form atomistic and molecular simulations of solid state, liq-uid, molecular, and biological systems. A description ofQUICKSTEP can be found in Ref. 37.
The GAPW method for the AE calculation of the NMRchemical shifts and the EPR g tensor was validated by com-parison, for a series of small isolated molecules, with theresults from conceptually similar gas-phase methods, !i" theNMR routines as implemented in the G03 program package38
!further referred to as the IGAIM-GA method", and !ii" the gtensor method of Schreckenbach and Ziegler34 as imple-mented in the ADF !Ref. 39" program package !further re-ferred to as the SZ method". Isolated molecules are approxi-mated by a supercell approach with a large cell size of!20 Å"3. We used a 300 Ry cutoff for the auxiliary planewave grid, the BLYP40,41 gradient-corrected exchange-correlation functional, and the efficient and numericallystable orbital transformation energy minimizer introduced inRef. 42. The Gaussian basis sets used in this work weretaken from the Environmental Molecular Sciences Labora-tory !EMSL" basis set exchange library.43
1. NMR
A comparison between the IGAIM method implementedin G03 !IGAIM-GA" and the presented IGAIM-GAPWmethod for the isotropic and the anisotropic chemical shiftsof a representative set of small molecules is shown in Fig. 1.Three different basis sets were used, namely, the Pople split-valence double-zeta plus polarization 6-31G!d , p",44,45 andthe augmented double and triple-zeta correlation consistentbasis sets aug-cc-pVDZ and aug-cc-pVTZ, respectively46–49.The test set is composed of 26 molecules, namely: C2H2,CH2O, CH3Cl, CH3F, CH4, Cl2, CO2, F2, FCl, H2O2, H2O,H2, H2S, HCl, HCN, HCOOH, HF, HNO3, N2O, N2, NH3,
NO2, O2, O3, PH3, and SO2. All the geometries were opti-mized at the BLYP/aug-cc-pVDZ level of theory.
For the small 6-31G!d , p" basis set, a maximal unsignedrelative error of 30% is observed for the isotropic chemicalshift of the carbon atom in HCN. The isotropic shift for thenitrogen atom in HCN is 3.9 ppm for IGAIM-GA and 5.1ppm for IGAIM-GAPW. This absolute error !1.2 ppm" is ofthe same order as for other nitrogen atoms with, e.g., 0.6ppm for NH3. Except for this extreme case, the other isotro-pic !anisotropic" chemical shifts have a smaller maximal un-signed relative error with, e.g., less than 1% !10%" for hy-drogen, 5% !0.5%" for carbon, and 2% !2%" for oxygen.
The unsigned relative error is greatly reduced for thelarger aug-cc-pVDZ and aug-cc-pVTZ basis sets. Theaug-cc-pVDZ provides a maximal unsigned relative error of5% !nitrogen in HCN" and 10% !H2" for, respectively, theisotropic and anisotropic components of the shift. The maxi-mal unsigned relative errors for the isotropic shift are 0.5%
-30
-20
-10
0
10
20
30
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/6-31G**
-12
-6
0
6
12
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/aug-cc-pVDZ
0 20 40 60 80test set
-4
-2
0
2
4
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/aug-cc-pVTZ
FIG. 1. Relative error of the IGAIM-GAPW method with respect to theIGAIM-GA method for the isotropic and anisotropic chemical shifts calcu-lated on a set of small molecules !see text for details". Note the differentscales in the three panels.
014106-5 Magnetic linear response properties J. Chem. Phys. 131, 014106 "2009#
Downloaded 27 Jan 2011 to 130.60.136.86. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
the Gaussian representation of the current density is also farfrom trivial. However, it generally holds that !gxy
SOO is arelatively small term in comparison with !gxy
ZKE and !gxySO.19
For this reason, we choose to neglect the contributions fromthe atom-centered current densities jxA
corr− j̃xAcorr to the G!0
components B̃x,G!0corr of the induced magnetic field. Being
computed from the soft current density j̃xcorr only, on the
reciprocal-space FFT-grid, these components will be soft too!hence the tilde". Bx,G=0
corr , on the other hand, will be computedanalytically via Eq. !9", using the Gaussian representation ofthe current density. In summary, !gxy
SOO is evaluated as fol-lows:
!gxySOO # 2$
FFTB̃xy,G!0
corr !r""̃s!r"d3r + 2%A$
UA
B̃xy,G!0corr !r"
#!"As !r" − "̃A
s !r""d3r + 2Bxy,G=0corr .
III. RESULTS AND DISCUSSION
A. Test calculations
All developments were implemented in QUICKSTEP,which is part of CP2K.36 CP2K is a freely available under theGNU General Public License !GPL" general program to per-form atomistic and molecular simulations of solid state, liq-uid, molecular, and biological systems. A description ofQUICKSTEP can be found in Ref. 37.
The GAPW method for the AE calculation of the NMRchemical shifts and the EPR g tensor was validated by com-parison, for a series of small isolated molecules, with theresults from conceptually similar gas-phase methods, !i" theNMR routines as implemented in the G03 program package38
!further referred to as the IGAIM-GA method", and !ii" the gtensor method of Schreckenbach and Ziegler34 as imple-mented in the ADF !Ref. 39" program package !further re-ferred to as the SZ method". Isolated molecules are approxi-mated by a supercell approach with a large cell size of!20 Å"3. We used a 300 Ry cutoff for the auxiliary planewave grid, the BLYP40,41 gradient-corrected exchange-correlation functional, and the efficient and numericallystable orbital transformation energy minimizer introduced inRef. 42. The Gaussian basis sets used in this work weretaken from the Environmental Molecular Sciences Labora-tory !EMSL" basis set exchange library.43
1. NMR
A comparison between the IGAIM method implementedin G03 !IGAIM-GA" and the presented IGAIM-GAPWmethod for the isotropic and the anisotropic chemical shiftsof a representative set of small molecules is shown in Fig. 1.Three different basis sets were used, namely, the Pople split-valence double-zeta plus polarization 6-31G!d , p",44,45 andthe augmented double and triple-zeta correlation consistentbasis sets aug-cc-pVDZ and aug-cc-pVTZ, respectively46–49.The test set is composed of 26 molecules, namely: C2H2,CH2O, CH3Cl, CH3F, CH4, Cl2, CO2, F2, FCl, H2O2, H2O,H2, H2S, HCl, HCN, HCOOH, HF, HNO3, N2O, N2, NH3,
NO2, O2, O3, PH3, and SO2. All the geometries were opti-mized at the BLYP/aug-cc-pVDZ level of theory.
For the small 6-31G!d , p" basis set, a maximal unsignedrelative error of 30% is observed for the isotropic chemicalshift of the carbon atom in HCN. The isotropic shift for thenitrogen atom in HCN is 3.9 ppm for IGAIM-GA and 5.1ppm for IGAIM-GAPW. This absolute error !1.2 ppm" is ofthe same order as for other nitrogen atoms with, e.g., 0.6ppm for NH3. Except for this extreme case, the other isotro-pic !anisotropic" chemical shifts have a smaller maximal un-signed relative error with, e.g., less than 1% !10%" for hy-drogen, 5% !0.5%" for carbon, and 2% !2%" for oxygen.
The unsigned relative error is greatly reduced for thelarger aug-cc-pVDZ and aug-cc-pVTZ basis sets. Theaug-cc-pVDZ provides a maximal unsigned relative error of5% !nitrogen in HCN" and 10% !H2" for, respectively, theisotropic and anisotropic components of the shift. The maxi-mal unsigned relative errors for the isotropic shift are 0.5%
-30
-20
-10
0
10
20
30
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/6-31G**
-12
-6
0
6
12
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/aug-cc-pVDZ
0 20 40 60 80test set
-4
-2
0
2
4
rela
tive
erro
rin
%
isotropicanisotropic
BLYP/aug-cc-pVTZ
FIG. 1. Relative error of the IGAIM-GAPW method with respect to theIGAIM-GA method for the isotropic and anisotropic chemical shifts calcu-lated on a set of small molecules !see text for details". Note the differentscales in the three panels.
014106-5 Magnetic linear response properties J. Chem. Phys. 131, 014106 "2009#
Downloaded 27 Jan 2011 to 130.60.136.86. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
Relative Error %
-4
4
0
00
-12
1230
-3020 20 60 80
20 40 60 8020 40 60 80
mal variation of about 9 ppm between the ISO and W3 cal-culations. Larger variations between the ISO and W3 calcu-lations are observed for the hydrogen and nitrogen atomswith 0.6 and 26 ppm, respectively. The last column of thetable contains the W5 calculations. The difference betweenthe W3 and W5 is very small with less than 1 ppm for thechemical shift of the carbons, 0.2 ppm for the hydrogens, and3 ppm for the nitrogens.
While the maximal variation in the chemical shifts withrespect to the size of the QM part is small with about 2 ppmfor carbon, 0.1 ppm for hydrogen, and 3 ppm for the nitrogenwithout any N–H bond !N1, N3, and N7", it can be morepronounced with up to 14 ppm for the other nitrogens !N6and N9". This finding reflects the formation of hydrogenbonds between the NH and NH2 groups with the solvent thatare not properly described in the W0 calculation.
1. g tensor of the E1! center in !-quartz
To further validate the method for the calculation of theg tensor and to apply it in a PBC simulation, we calculatedthe g tensor of the positively charged oxygen vacancy E1!center in !-quartz. Being one of the most abundant pointdefects in silica and due to its importance in the degradationof the performance of the gate oxide in transistors, the E1!center has been the subject of a large number of calculations!an excellent overview can be found in Ref. 56". To ourknowledge, the g tensor of this defect has been calculatedonly once,19 using the GIPAW g-tensor method. However,the !periodic" simulation cell that was used included only 71
TABLE IV. Calculated 13C, 1H, and 15N chemical shifts of an isolated adenine !upper part of the table". Valuesin parentheses are corrected for core electrons !see text for details". Absolute chemical shifts of the referencesare shown in the lower part of the table. The subscripts t, n, b, py, pr, and a refer to TMS, MeNO2, benzene,pyridine, pyrrole, and aniline, respectively. All the values are in ppm.
IGAIM-GAPWa IGAIM-GAa IGAIM-GAb CSGT-CPc IGAIM-GAPWd GIPAWe
C2 164 164 166 136 !168" 162 166C4 159 159 162 130 !161" 156 162C5 128 128 130 93 !124" 126 131C6 163 163 165 134 !166" 161 164C8 141 141 143 115 !146" 140 143H2 8.4 8.4 8.5 7.4 8.4 8.5H8 7.5 7.5 7.7 7.0 7.6 8.1N1 "134 "134 "135 "143 !"118" "128 "132N3 "142 "142 "146 "152 !"127" "136 "141N6 "319 "319 "327 "285 !"332" "310 "322N7 "129 "129 "132 "154 !"129" "123 "128N9 "235 "235 "240 "217 !"240" "226 "234Ct 177 177 175 7 181 184Ht 31.3 31.3 31.3 30.6 31.1 30.9Nn "159 "159 "166 "299 "152 "139Cb 37 "99Npy "119 "227Npr 77 "79Na 171 "8
aBLYP/cc-pVQZ.bBLYP/aug-cc-pV5Z.cBLYP/200 Ry.dPBE/cc-pVQZ.ePBE/100 Ry.
TABLE V. Calculated 13C, 1H, and 15N chemical shifts of adenine at theBLYP/cc-pVQZ level of theory within a QM/MM framework !see text fordetails". All the values are in ppm.
ISO W0 W3 W5
C2 164 164 166 166C4 148 148 148 148C5 120 120 120 120C6 160 160 159 159C8 145 152 154 154H2 7.8 7.8 7.8 7.9H8 7.7 8.3 8.2 8.4N1 "115 "129 "128 "125N3 "127 "147 "144 "145N6 "330 "330 "318 "317N7 "121 "144 "147 "149N9 "249 "237 "226 "223
FIG. 3. Labeling scheme for the adenine.
014106-8 Weber et al. J. Chem. Phys. 131, 014106 !2009"
Downloaded 27 Jan 2011 to 130.60.136.86. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
Isolated Adenine
41
external reference
TMSMeNO2
pyrid.benz.
alinine
mal variation of about 9 ppm between the ISO and W3 cal-culations. Larger variations between the ISO and W3 calcu-lations are observed for the hydrogen and nitrogen atomswith 0.6 and 26 ppm, respectively. The last column of thetable contains the W5 calculations. The difference betweenthe W3 and W5 is very small with less than 1 ppm for thechemical shift of the carbons, 0.2 ppm for the hydrogens, and3 ppm for the nitrogens.
While the maximal variation in the chemical shifts withrespect to the size of the QM part is small with about 2 ppmfor carbon, 0.1 ppm for hydrogen, and 3 ppm for the nitrogenwithout any N–H bond !N1, N3, and N7", it can be morepronounced with up to 14 ppm for the other nitrogens !N6and N9". This finding reflects the formation of hydrogenbonds between the NH and NH2 groups with the solvent thatare not properly described in the W0 calculation.
1. g tensor of the E1! center in !-quartz
To further validate the method for the calculation of theg tensor and to apply it in a PBC simulation, we calculatedthe g tensor of the positively charged oxygen vacancy E1!center in !-quartz. Being one of the most abundant pointdefects in silica and due to its importance in the degradationof the performance of the gate oxide in transistors, the E1!center has been the subject of a large number of calculations!an excellent overview can be found in Ref. 56". To ourknowledge, the g tensor of this defect has been calculatedonly once,19 using the GIPAW g-tensor method. However,the !periodic" simulation cell that was used included only 71
TABLE IV. Calculated 13C, 1H, and 15N chemical shifts of an isolated adenine !upper part of the table". Valuesin parentheses are corrected for core electrons !see text for details". Absolute chemical shifts of the referencesare shown in the lower part of the table. The subscripts t, n, b, py, pr, and a refer to TMS, MeNO2, benzene,pyridine, pyrrole, and aniline, respectively. All the values are in ppm.
IGAIM-GAPWa IGAIM-GAa IGAIM-GAb CSGT-CPc IGAIM-GAPWd GIPAWe
C2 164 164 166 136 !168" 162 166C4 159 159 162 130 !161" 156 162C5 128 128 130 93 !124" 126 131C6 163 163 165 134 !166" 161 164C8 141 141 143 115 !146" 140 143H2 8.4 8.4 8.5 7.4 8.4 8.5H8 7.5 7.5 7.7 7.0 7.6 8.1N1 "134 "134 "135 "143 !"118" "128 "132N3 "142 "142 "146 "152 !"127" "136 "141N6 "319 "319 "327 "285 !"332" "310 "322N7 "129 "129 "132 "154 !"129" "123 "128N9 "235 "235 "240 "217 !"240" "226 "234Ct 177 177 175 7 181 184Ht 31.3 31.3 31.3 30.6 31.1 30.9Nn "159 "159 "166 "299 "152 "139Cb 37 "99Npy "119 "227Npr 77 "79Na 171 "8
aBLYP/cc-pVQZ.bBLYP/aug-cc-pV5Z.cBLYP/200 Ry.dPBE/cc-pVQZ.ePBE/100 Ry.
TABLE V. Calculated 13C, 1H, and 15N chemical shifts of adenine at theBLYP/cc-pVQZ level of theory within a QM/MM framework !see text fordetails". All the values are in ppm.
ISO W0 W3 W5
C2 164 164 166 166C4 148 148 148 148C5 120 120 120 120C6 160 160 159 159C8 145 152 154 154H2 7.8 7.8 7.8 7.9H8 7.7 8.3 8.2 8.4N1 "115 "129 "128 "125N3 "127 "147 "144 "145N6 "330 "330 "318 "317N7 "121 "144 "147 "149N9 "249 "237 "226 "223
FIG. 3. Labeling scheme for the adenine.
014106-8 Weber et al. J. Chem. Phys. 131, 014106 !2009"
Downloaded 27 Jan 2011 to 130.60.136.86. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
CPMD CSGT
BLYP/200Ry
G03 IGAIM
BLYP /cc-pVQZ BLYP /aug-cc-pVZ5
CP2K IGAIM
BLYP/cc-p-VQZ
Geometry optimized in gas phase : BLYP/6-31G(dp)
PP corrected
pyrrol
PP correction
Hydrated Adenine
42
mal variation of about 9 ppm between the ISO and W3 cal-culations. Larger variations between the ISO and W3 calcu-lations are observed for the hydrogen and nitrogen atomswith 0.6 and 26 ppm, respectively. The last column of thetable contains the W5 calculations. The difference betweenthe W3 and W5 is very small with less than 1 ppm for thechemical shift of the carbons, 0.2 ppm for the hydrogens, and3 ppm for the nitrogens.
While the maximal variation in the chemical shifts withrespect to the size of the QM part is small with about 2 ppmfor carbon, 0.1 ppm for hydrogen, and 3 ppm for the nitrogenwithout any N–H bond !N1, N3, and N7", it can be morepronounced with up to 14 ppm for the other nitrogens !N6and N9". This finding reflects the formation of hydrogenbonds between the NH and NH2 groups with the solvent thatare not properly described in the W0 calculation.
1. g tensor of the E1! center in !-quartz
To further validate the method for the calculation of theg tensor and to apply it in a PBC simulation, we calculatedthe g tensor of the positively charged oxygen vacancy E1!center in !-quartz. Being one of the most abundant pointdefects in silica and due to its importance in the degradationof the performance of the gate oxide in transistors, the E1!center has been the subject of a large number of calculations!an excellent overview can be found in Ref. 56". To ourknowledge, the g tensor of this defect has been calculatedonly once,19 using the GIPAW g-tensor method. However,the !periodic" simulation cell that was used included only 71
TABLE IV. Calculated 13C, 1H, and 15N chemical shifts of an isolated adenine !upper part of the table". Valuesin parentheses are corrected for core electrons !see text for details". Absolute chemical shifts of the referencesare shown in the lower part of the table. The subscripts t, n, b, py, pr, and a refer to TMS, MeNO2, benzene,pyridine, pyrrole, and aniline, respectively. All the values are in ppm.
IGAIM-GAPWa IGAIM-GAa IGAIM-GAb CSGT-CPc IGAIM-GAPWd GIPAWe
C2 164 164 166 136 !168" 162 166C4 159 159 162 130 !161" 156 162C5 128 128 130 93 !124" 126 131C6 163 163 165 134 !166" 161 164C8 141 141 143 115 !146" 140 143H2 8.4 8.4 8.5 7.4 8.4 8.5H8 7.5 7.5 7.7 7.0 7.6 8.1N1 "134 "134 "135 "143 !"118" "128 "132N3 "142 "142 "146 "152 !"127" "136 "141N6 "319 "319 "327 "285 !"332" "310 "322N7 "129 "129 "132 "154 !"129" "123 "128N9 "235 "235 "240 "217 !"240" "226 "234Ct 177 177 175 7 181 184Ht 31.3 31.3 31.3 30.6 31.1 30.9Nn "159 "159 "166 "299 "152 "139Cb 37 "99Npy "119 "227Npr 77 "79Na 171 "8
aBLYP/cc-pVQZ.bBLYP/aug-cc-pV5Z.cBLYP/200 Ry.dPBE/cc-pVQZ.ePBE/100 Ry.
TABLE V. Calculated 13C, 1H, and 15N chemical shifts of adenine at theBLYP/cc-pVQZ level of theory within a QM/MM framework !see text fordetails". All the values are in ppm.
ISO W0 W3 W5
C2 164 164 166 166C4 148 148 148 148C5 120 120 120 120C6 160 160 159 159C8 145 152 154 154H2 7.8 7.8 7.8 7.9H8 7.7 8.3 8.2 8.4N1 "115 "129 "128 "125N3 "127 "147 "144 "145N6 "330 "330 "318 "317N7 "121 "144 "147 "149N9 "249 "237 "226 "223
FIG. 3. Labeling scheme for the adenine.
014106-8 Weber et al. J. Chem. Phys. 131, 014106 !2009"
Downloaded 27 Jan 2011 to 130.60.136.86. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/about/rights_and_permissions
QM/MM with 827 H2O: snapshot from classical MD
CP2K IGAIM
BLYP/cc-p-VQZ
QM water for proper description of solute/solvent
h-bonds